摘要
在基于HOG特征的SVM行人检测算法的基础上,提出了组合分类器的改进算法。该算法首先采用多尺度滑动窗口提取HOG特征,并对单个SVM分别进行训练,再将训练好的SVM分别采用串联、并联结构形成新分类器后对行人进行检测。为解决用多尺度滑动窗口提取特征时产生的目标候选区域重叠问题,采用非极大值抑制算法对重叠区域进行融合,进而得到准确候选区。实验表明,组合的SVM分类器可以有效降低误检率和漏检率。
On the basis of histogram of oriented gradient and support vector machine(HOG-SVM)algorithm,this paper proposed an improved algorithm for combination classifiers.Firstly,This algorithm uses multi-scale sliding windows to extract the HOG features and trains SVM separately.Then,the trained SVM which is formed to a new classifier in series or parallel is used to detect pedestrian.In order to solve the problem that the target area is overlapped when features are extracted in multi-scale sliding windows,the non-maximum suppression(NMS)algorithm is used to fuse the rectangles and to get exact candidate region.Experiments show that combined SVM classifiers can effectively reduce the false detection rate and missed rate.
出处
《计算机科学》
CSCD
北大核心
2017年第S1期188-191,共4页
Computer Science
基金
国家自然科学基金资助项目(61103136)
武汉工程大学创新基金资助项目(CX2015057)资助
关键词
行人检测
HOG
SVM
NMS
组合分类器
Pedestrian detection
Histogram of oriented gradient(HOG)
Support vector machine(SVM)
Non-maximum suppression(NMS)
Combination classifiers
